Statistical methodology for clinical trials of caries prophylactic agents--current knowledge.

Statistical methodology underlying clinical trials of caries prophylactic agents published up to 1982 is summarized. The various caries indices used to express an individual's caries increment over a period of time are considered, along with some standard models used to describe caries incremental and prevalence data. The parametric statistical techniques generally employed to test for significant differences between the caries preventive agents under investigation are robust for the departures from normality found in caries data with the large sample sizes involved. Methods of estimating sample sizes are considered along with the various multiple comparison tests suitable for analysing multi-group trials. The advantages of using either an analysis of covariance or stratification procedures in certain situations are discussed, along with suitable covariables which have been put forward. Ways of improving the efficiency of clinical trials by selecting subjects are also considered. Models put forward to describe diagnostic errors are examined. However, it is generally found that examiner error in long term trials is small. The use of sequential tests in clinical trials needs further development.